Mastery data from 14343 different summoners from Iron to Diamond. For each summoner, I took their top 3 champions (three champions with the most mastery points) as their three mains. I then looked at how often certain champions co-occured (were correlated) in mains across players.
I first wanted to eyeball every champions relationship to all other champions as a giant correlation matrix (heatmap) just to see if there is any meaningful relationships at all in what champions are often mained by the same players. This checkerboard visual is showing every pairwise relationship between all champions. A blue square means those two champions (line up the champ name on the vertical and horizontal axis) are often mained together. A red squares means those two champions are rarely mained together. I ordered the champions using hierarchical clustering which creates these “groups” of champions based on their similarity (how often they are mained together across players) to other champions which creates several large blue clusters. For example, in the very bottom left you can see an assassin cluster emerging where if you play one assassin you are likely to play another. I know it’s hard to see so let’s zoom in a bit.
Upon zooming in, we can see a support-like, mage cluster and an assassin-like cluster forming (blue clusters). This shows how players that play mage/support champs also play many similar champions, but less often play assassins (red colors indicate negative correlation)
Champions with the most mastery points (most popular)
## # A tibble: 6 x 2
## name mean_points
## <chr> <dbl>
## 1 Yasuo 35018.
## 2 Lux 28937.
## 3 Kaisa 28097.
## 4 Kayn 23944.
## 5 Jhin 23916.
## 6 Ezreal 23361.
I wrote a function where you input a champion’s name and it spits out the top 3 champs players most often and least often mained with that champion. Using the results from this, I made a graphic (photoshop) of 8 example champions and their most likely and least likely mains.
## [1] "Teemo"
## most_mained_together least_mained_together
## 1 Heimerdinger Kaisa
## 2 Garen Yasuo
## 3 Singed Graves
Another way to visualize this data is as a network. I visualized only champions that had strong connections with other champions (were often mained together). Connections (white lines) indicate a strong relationship in how often champions are mained together across players. Because the network is force-directed, these connections, but also distance, indicate how likely those champions are to be mained together. You can see communities that naturally emerge that seem to reflect champion playstyles (assassins cluster, support cluster, etc.). You can also see some neat connections between these communities such as the champions with pulls (Thresh/Blitz/Pyke) being frequently mained together. But with Pyke having more direct connections with the assassin cluster.